Photonic IC Enabled

Coherent Optical Systems

White Paper

NeoPhotonics - White Paper 1

Photonic ICs and modules set the foundation
for telecom infrastructure

2
Table of Contents

1. Introduction to coherent transmission systems 4

2. How a coherent transmission system works 4

a. Coherent Detection 4

b. Modulation Schemes 6

c. Building Blocks 8

3. Advantages of 100G coherent systems over conventional 10G/40G systems 10

4. Future Trend of Coherent Systems and Components 10

a. Emerging 400G coherent systems and the required optoelectronic components 10

b. Pluggable coherent modules and the required optoelectronic components 13

5. Why NeoPhotonics? 15

6. Summary 15

7. References 17
1. Introduction most important elements for telecom transmission
infrastructure in the next decade, NeoPhotonics
has focused its research and development activities
on developing all the enabling PICs, modules, and
Global internet protocol (IP) traffic will grow 3-fold analog electronic ICs.
from 2015 to 20201. To transport the explosive
Internet traffic, a corresponding bandwidth growth
in long-haul and metro optical fiber networks will
have to happen at the same time. Consequently,
2. How a 100Gb/s coherent
optical fibers in service providers’ networks have system works
been quickly exhausted. In order to slow down the
consumption rate of precious optical fiber trunk
lines, service providers are in urgent need of more
bandwidth-efficient transmission systems. To that
2a. Coherent Detection
end, newly developed 100 Gb/s coherent systems
The key feature of a “coherent” system is charac-
(more description of coherent systems is given in
terized by its capability to do “coherent detection,”
Sec.2) have become a timely savior because they
which means that a receiver can track the phase of
can provide ten times more transport capacity than
a transmitter (and hence have “phase coherence”) so
today’s 10Gb/s systems, while providing a much
as to extract any phase and frequency information
higher spectral efficiency and simpler transmission
carried by a transmitted signal.
infrastructure.

Coherent detection is well known in wireless com-

Further motivation for service providers to use 100
munication systems. In those wireless systems, a
Gb/s coherent systems is because not only are 100
radio frequency (RF) local oscillator (LO) is tuned to
Gb/s Ethernet (100GE) and OTN (Optical Trans-
“heterodyne” (“heterodyne” is a signal processing
port Network) interfaces well defined in existing
technique which combine a high-frequency signal
standards (IEEE802.3ba2 and ITU-T3, respectively),
@ f1 with another @ f2 to produce a lower frequen-
but the modulation and detection technologies
cy signal @ f1 - f2) with a received signal through
are also defined in OIF (Optical Inter-networking
an RF mixer, as shown in Fig.1(a), so that both the
Forum) implementation agreements4. As a result of
amplitude and phase information contained in an
the well-defined standards, 100 Gb/s technology
RF carrier can be recovered in the following digital
adoption is being driven by products in all network
signal processor (DSP).
layers, from transport systems to edge routers
and datacenter switches. Therefore, we expect to
For an optical coherent system, shown in Fig. 1(b),
witness a sea change in optical transport networks
a narrow linewidth tunable laser, serving as an LO,
that will become “coherent-centric” in the next few
tunes its frequency to “intradyne” with a received
years.
signal frequency through an optical coherent mixer,
and thereby recovers both the amplitude and phase
NeoPhotonics is a fast-growing optical compo-
information contained in a particular optical carrier.
nents and sub-system company, and is a leading
Here, “intradyne” means that the frequency differ-
supplier of 100 Gb/s coherent optical components.
ence between an LO and a received optical carrier is
Knowing that photonic ICs (Integrated Circuits),
close to, but does not have to be zero.
or PICs, set the foundation for worldwide telecom
infrastructure (as is illustrated in the first diagram
This implies that the frequency and phase of an LO
of this White Paper), NeoPhotonics started with
do not have to be actively controlled to an extreme
silica PLC (Planar Lightwave Circuit) components,
accuracy, therefore avoiding the use of a complicat-
and acquired complementary indium-phosphide and
ed optical phase locked loop.
gallium-arsenide platforms. Furthermore, envi-
sioning that coherent systems are becoming the

4
Fig.1

A simple illustration of how coherent detection works for

(a) wireless systems and (b) optical coherent systems.

NeoPhotonics - White Paper 5

The contrast to coherent detection is direct detec-
tion, typically used by 10Gb/s or lower-speed. In
a direct detection receiver, its photo-detector only
responds to changes in the received signal optical
power, and cannot extract any phase or frequency
information from the optical carrier.
Coherent detection therefore offers several key
advantages compared to direct detection:
(1) Greatly improved receiver sensitivity.
(2) Can extract amplitude, frequency, and
phase information from an optical carrier, and con-
sequently achieve much higher capacity in the same
spectral bandwidth.
(3) Its DSP can compensate very large
chromatic and polarization mode dispersion caused
by optical fibers, and eliminate the need of optical
dispersion compensators and the associated optical
amplifiers. This saves not only significant capex, but
also simplifies optical network design tremendously
because the complicated dispersion map associated
with 10G and 40G direct detection systems is no
longer needed.
(4) When using balanced detectors with a
high common mode noise rejection ratio (CMRR),
not only can signal-to-noise ratio (SNR) be improved
further, but also agile wavelength selection can be
achieved by LO tuning without the use of an optical
filter or demultiplexer.
It should also be noted that laser phase noise of an
LO is an important impairment in coherent systems
as it impacts the “phase coherence.” The “linewidth”
parameter of a laser diode is directly related to its
phase noise.
2b. Modulation Scheme
A modulation methodology allows information to
be carried by an optical or radio-frequency wave. A
few modulation schemes have been used for 100
Gb/s coherent systems, including DP-BPSK (mainly
used for undersea cable systems), DP-QPSK (mainly
used for long-haul terrestrial systems), and DP-
16QAM (mainly used for metro/regional terrestrial
systems). Here “DP” stands for “dual-polarization,”
“BPSK” stands for “binary phase shift keying,”
6
“QPSK” stands for “quadrature phase shift keying,”
and “16QAM” stands for “16 quadrature-ampli-
tude-modulation.” The details of each term will be
explained as follows.
First of all, we notice that dual polarizations are
used in all modulation schemes above. A simple
analogy can be used to explain why dual polariza-
tions are used to carry twice as much information.
As shown in Fig. 2, a 3D movie theater uses two
projectors to generate two orthogonal polariza-
tions on a silver screen, and the reflected light
which combines the two polarizations is received
by a viewer who is wearing polarized glasses.
Each polarization is received by only one eye, and
consequently enables the perception of depth. A
100Gb/s coherent transmission system follows the
same principle that its optical transmitter gener-
ates two orthogonal polarizations, each carrying
a stream of 50Gb/s of data, and its integrated
coherent receiver (ICR) and an optical local oscilla-
tor (LO) function as the polarized glasses, while a
high-speed digital signal processor (DSP) functions
as human eyes and brain.
Next, let us see how BPSK and QPSK work. BPSK
can be understood by comparing it to BASK (Bi-
nary Amplitude Shift Keying), which is commonly
used in today’s conventional 10 Gb/s optical sys-
tems. BASK modulates data via “light-off,” repre-
senting a symbol “0,” and “light-on,” representing a
symbol “1,” which result in merely 1 bit per symbol.
In contrast, BPSK uses phases 0 and π to transport
the same amount of information as BASK. QPSK
uses four phases, e.g., 0, π/2, π, and 3π/2 to trans-
port twice as much information as BASK or BPSK,
and achieves 2 bits per symbol. As a result, QPSK
in combination with dual polarizations (i.e., DP-
QPSK) can achieve 2×2=4 bits per symbol. This
implies that a 100Gb/s DP-QPSK coherent optical
link only requires 25G symbols/s or 25G baud,
which in turn reduces the bandwidth requirement
of all electrical circuits, printed circuit boards, and
optical components by four times. Similarly, the
optical spectrum that must be allocated to a DP-
QPSK 100 Gb/s signal can be nominally four times
less than if one were to use an “on-off” 100Gb/s
optical signal. In summary, in comparison to
Fig.2

A 3D theater viewing, analogous to a dual-polarization

coherent transmission system, is using dual-polarization
projectors at the transmitter, and a polarized glasses at
the receiver (http://www.cinema3dglass.com).

NeoPhotonics - White Paper 7

10Gb/s BASK modulation which has a spectral effi- Fig.3
ciency of 1 bit/s/symbol, 100Gb/s DP-QPSK offers
: symbols
four times improvement (4 bits/s/symbol).

2 bits/symbol
A DP-16QAM signal contains not only dual
(01)
polarizations, but also both amplitude and phase
modulation from symbol to symbol, and therefore
further improves the spectral efficiency of DP-
QPSK by two times, i.e., it has a spectral efficiency (11) (00)

of 8 bits/s/symbol. A simple illustration is shown

in Fig.3, where we can see that to change from one
symbol to another in a QPSK constellation diagram (10)
requires only phase variation, while to change one
symbol to another in a DP-16QAM constellation 4 bits/symbol
diagram requires both phase and amplitude varia-
tion. We also notice that each symbol (running at
25G symbols/s) in a QPSK constellation diagram (0000) (0100) (1100) (1000)

is composed of 2 bits, therefore we obtain 2 bits/

symbol × 25G symbols/s= 50 Gb/s in a single (0001) (0101) (1101) (1001)

polarization and 100 Gb/s with two polarizations.

Similarly, each symbol in a 16QAM constellation
(also running at 25G baud) is composed of 4 bits, (0011) (0111) (1111) (1011)
therefore we obtain 4 bits/symbol × 25G symbols/
s= 100 Gb/s in a single polarization, and 200Gb/s (0010) (0110) (1110) (1010)
in two polarizations.

2c. Building Blocks The constellation diagrams of a QPSK and 16QAM

signals, respectively. For QPSK, we have 4 symbols
Having reviewed the coherent modulation and with 2 bits/symbol, while for 16QAM, we have 16
detection schemes, we can now take a look at symbols with 4 bits /symbol. Each symbol is as-
the basic building blocks of the current coherent sumed to be running at ≥ 25Gbymbols/s.
transmitters and receivers, as shown in Fig.4. The
key components at the transmitter side are: an
integrated tunable laser assembly (ITLA), a lithi-
um-niobate, silicon, or indium-phosphide I/Q (In-
and Quadrature-phase) modulator, and four driver
amplifiers. The key components at the receiver
side are: an integrated coherent receiver (ICR), an
ITLA serving as an LO, and a DSP chip. The DSP is
used to separate the two polarizations, compen-
sate various transmission impairments, and recover
carrier frequency and phase.

Gen2 coherent transmitters and receivers, to be

described in Sec.5, require another level of photon-
ic and electronic integration to get to smaller size
and lower power consumption to fit inside plugga-
ble coherent transceivers.

8
Fig.4

The basic building blocks of coherent

transmitter and receiver.

NeoPhotonics - White Paper 9

3. Advantages of 100G coher-
ent systems over conventional
10G/40G systems
As mentioned in Sec.1, 100G has been standard-
ized at all layers which has resulted in concentrated
efforts on the development of optical and electron-
ic components, thus enabling a high-momentum
ecosystem. In contrast, 40G is and will remain a
fragmented market because there are still four
modulation formats competing against each other
for long-haul DWDM transport systems, including
coherent, DQPSK, DPSK, and ODB (optical dual
binary). As a result, the possible range of cost re-
duction for 100G DP-QPSK transponders has been
much steeper than that of 40G transponders.
When comparing 10G to 100G for long-haul trans-
mission systems, due to the fact that 100G offers at
least a ten-fold capacity increase, and saves signif-
icant system cost by eliminating numerous optical
dispersion compensators (by using the powerful
DSP in a 100G coherent receiver) and the associat-
ed optical amplifiers (which were needed to com-
pensate for the loss of those dispersion compensa-
tors), 10G transponders simply cannot compete in
this market.
This is especially true when we consider the fact
that a 100G transponder module already offers a
lower price than ten tunable 10G transponders in
XFP or SFP+ form factors.
4. Future Trends of Coherent
Systems and Components
There are three future trends of coherent sys-
tems, i.e., higher baud rate (evolving from today’s
32Gbaud to tomorrow’s 45, 56, and 64Gbaud),
higher constellations (evolving from today’s DP-
QPSK, DP-16QAM, to tomorrow’s DP-32QAM and
DP-64QAM), and smaller form factors (evolving
from today’s line-cards and CFP modules to tomor-
10
row’s CFP2, CFP4, or even QSFP modules). These
three trends are illustrated in Fig. 5. Also shown in
the figure are the associated component trends (in
green arrows) which are needed to enable each of
the coherent system trends: For higher baud rates,
higher bandwidth transmitters and receivers are
needed (evolving from today’s ~20GHz components
to tomorrow’s ~30 or 40GHz components); for
higher constellations, the transmitting laser and lo-
cal oscillator laser both need to have a narrow-line-
width (e.g., < 100KHz); for higher port density and
lower power consumption, the components needed
to enable small form factors are micro- or nano-tun-
able lasers, modulators and receivers.
4.a Emerging 400G coherent systems and the re-
quired optoelectronic components
Quite a number of 400G coherent systems products
have been deployed. Here we discuss the pros and
cons of different 400G modulation techniques.
Starting 2015: 2×200Gb/s using 32Gbaud/DP-
16QAM (Fig.6(a))
As mentioned in Sec.2b, a wavelength could double
its capacity from 100Gb/s to 200Gb/s by replacing
DP-QPSK with DP-16QAM modulation, while the
baud rate is maintained at 32Gbaud. The same
hardware in Fig.4 can be used, except that four
limiting driver amplifiers need to be replaced by
four linear driver amplifiers. As a result, the same
100G optical components can continue to be used
while the capacity is doubled, thereby achieving
essentially 50% cost reduction. By using two such
wavelengths, we can obtain 2×200Gb/s or 400Gb/s
capacity.
A transmitter DSP/DAC must be used, not only
for pulse-shaping, but also for assembling bits
into symbols. The total occupied spectrum of
2×200Gb/s DP-16QAM is ~75 GHz. The penalty
of 2×200Gb/s, however, is that its transmission
distance of 400~600km is much shorter than the
2,500~3,000km of a 100Gb/s DP-QPSK signal.
This is due to the fact that the 16QAM symbol con-
stellations are much closer than those of QPSK (see
Fig. 3), and therefore cause 7~8 dB OSNR penalty.
Fig.5

Three future trends for coherent systems

and the associated components.

NeoPhotonics - White Paper 11

Fig.6 Four potential configurations of 400G optical transponder. PC: power coupler; DSP: digital signal pro-
cessing; DAC: digital-to-analog converter; ADC: analog-to-digital converter. The transmission distances are
first-order estimate and depend on inter-office distances, fiber types, and optical amplifier types.

12
Starting 2016: 2×200Gb/s using 45Gbaud/DP-
8QAM (Fig.6(b))
By upgrading the speed of DACs, drivers, modula-
tors, ICRs, and ADCs, one can use two wavelengths
to achieve 2x200Gb/s capacity via 45Gbaud/DP-
8QAM modulation. The reason to change from
32Gbaud/DP-16QAM to 45Gbaud/DP-8QAM is
that the latter can achieve a longer transmission dis-
tance (~1,500km) due to the lower constellations.
Starting 2017: 2×200Gb/s using 64Gbaud/DP-
QPSK (Fig.6(c))
coherent receiver are used, the cost per bit can be
significantly lower than a conventional 100Gb/s
coherent transceiver today.
4.b Pluggable coherent modules and
the required optoelectronic components
Pluggable coherent optical transceiver modules are
needed mainly because of the highly desirable fea-
tures such as “pay-as-you-grow” and easy replace-
ment, which a multiple-port line-card cannot offer.
It is essentially following the same deployment
history of 10G pluggable transceivers.

Similar to 45Gbaud/DP-8QAM, this method uses There are two main categories of pluggable co-
an even higher baud rate while lowering the con- herent transceivers: analog coherent optics (ACO)
stellations further to QPSK. It can also achieve a and digital coherent optics (DCO). ACO implies
much longer transmission distance than 32Gbaud/ that the DSP ASIC is outside the transceiver mod-
DP-16QAM. However, the speed of all components ule, while DCO implies that DSP ASIC is inside the
(DACs, drivers, modulators, ICRs, and ADCs) needs transceiver module. ACO and DCO can fit inside a
to be upgraded. form factor of CFP, CFP2, CFP4, or even QSFP. Fig.
7 (next page) illustrates the electronic and optical
Starting 2017/2018: 1×400Gb/s using 56~64G components inside a CFP-DCO, a CFP2-ACO, and
baud/DP-16QAM (Fig.6(d)) a CFP2-DCO. The key electronic components are
DSP ASIC (which includes digital-to-analog convert-
Data center interconnection is one of the fastest er (DAC) and analog-to-digital converter (ADC)) and
growing areas among all optical fiber networks. four driver amplifiers. The key optical components
The typical interconnection distance is shorter than are micro-integrated tunable laser assembly (μITLA),
80km. As a result, the coherent system design is micro-modulator (μMOD), and micro-integrated
very different from that of a long-haul system. For coherent receiver (μICR).
this application, one can try to maximize the baud
rate and constellations such that the OSNR can
still remain beyond 30~32dB, which is a lot looser
requirement than that for typical metro-core or
long-haul systems. However, higher constellations
requires high resolution DACs and ADCs (i.e., high
effective number of bits or ENOBs) to maintain a
sufficient signal-to-noise ratio, which will not be
available until late 2016/early 2017. Therefore, in
the short term, to develop a single-carrier 400Gb/s,
most companies will use a high baud rate rather
than a high constellation. For example, a single
wavelength of 56~64Gbaud/DP-16QAM can offer
a total capacity of 448~512Gb/s.

In a single-wavelength 400Gb/s coherent trans-

ceiver module, since only one tunable laser (serving
both as TX and LO), one I/Q modulator, and one

NeoPhotonics - White Paper 13

Fig 7 Illustration of CPF-DCO, CFP2-ACO, and CFP2-DCO

CFP-DCO (long-haul)

(+) Plug into any client-side

line-card with CFP slots
(--) Poor port density

CFP2-ACO (metro)

(+) Good port density

(--) Analog RF interface

CFP2-DCO (metro)

(+) Good port density

(+) Plug into any client-side
line-card with CFP2 slots
(--) Mechanically (thermal &
size) challenging

14
Table 1. Market availability and application areas
Regarding the optical components, we believe that
the best approach to keep the pluggable transceiver
cost low is to leverage the existing high-volume of
µITLA and µICR which have been used for line-card
applications. Alternatively, when either one of the
two devices is combined with a low-volume µMOD
in a new gold-box, the total cost of the transceiver
module will be increased significantly due to the low
yield and low volume of the new gold-box. Table 1
above summarizes the application areas and esti-
mated timeframe for each type of pluggable coher-
ent transceiver. Note that CFP4 may be completely
skipped, depending on how soon QSFP-like high
port-density modules will be available.
5. Why NeoPhotonics?
NeoPhotonics has been a market share leader in
both the ICR and narrow linewidth tunable lasers
and has been a leader in developing next generation
component technologies. These 100G and beyond
products utilize NeoPhotonics Advanced Hybrid
Photonic Integration technology, which consists of
photonic integrated circuits (PICs) that comprise
both arrayed and individual photonic functional
elements using optimized materials systems and
processes from NeoPhotonics in-house Silicon, Indi-
um Phosphide and Gallium Arsenide wafer fabrica-
tion facilities. These individual PICs from different
materials are then combined using hybrid integra-
tion technology to make complete products, such
as Integrated Coherent Receiver (ICR) for 100G
NeoPhotonics - White Paper
coherent transport applications. The complete suite
of coherent-related photonic and electronic com-
ponents offered by NeoPhotonics is summarized in
Table 2 (next page).
6. Summary
Although its fundamental concept is derived from
wireless communications systems, coherent detec-
tion has started another paradigm shift in optical
fiber communications. Its impact has become as
large as that which commercial laser diodes and er-
bium-doped fiber amplifiers brought to the industry.
Consequently, long-haul and metro optical net-
works are bound to become coherent-centric in the
next decade. Infonetics projects that the number of
100G ports deployed will grow with a 5 year CAGR
of 47%, reaching more than 700,000 in 20196.
In addition, by the time pluggable coherent tran-
sponders such as CFP2 are available, metro and
Ethernet-over-DWDM markets will both be rapidly
expanding, and the potential volume could be 2~3
times higher - based on the experience the industry
had in 10Gb/s market. NeoPhotonics has already
been the market share leader today in ICR and nar-
row linewidth tunable lasers for the current coher-
ent components, and its leading PIC technology will
continue to enable the next-generation coherent
components and transceiver modules.
15
Table 2 Extending Our PIC-Based Coherent Product Portfo-
lio into More Powerful and Compact form factors

Our PIC based Coherent Solutions enable

400G transport for Long Haul, Metro
and Inter-Datacenter Networks

NeoPhotonics next-generation of coherent products

16
References

1. http://www.cisco.com/go/vni

2. IEEE 802.3ba, Media Access Control Parameters, Physical Layers and Management Parameters for
40 Gb/s and 100 Gb/s Operation

3. http://www.itu.int/rec/T-REC-G.709/en (Recommendation ITU-T G.709/Y.1331, 02/2012)

4. http://www.oiforum.com/public/documents/OIF-FD-100G-DWDM-01.0.pdf

5. HIS-Infonetics “100G+ Coherent Optical Equipment Ports Worldwide Market Size and Forecast” Au-
gust 2015.

NeoPhotonics - White Paper 17

18